A game-theoretic analysis of DoS attacks on driverless vehicles

Abstract

Driverless vehicles are expected to form the foundation of future connected transport infrastructure. A key weakness of connected vehicles is their vulnerability to physical-proximity attacks such as sensor saturation attacks. It is natural to study whether such attacks can be used to disrupt swarms of autonomous vehicles used as part of a large fleet providing taxi and courier delivery services. In this paper, we start to examine the strategic options available to attackers and defenders (autonomous-fleet operators) in such conflicts. We find that attackers have the upper hand in most cases and are able to carry out crippling denial-of-service attacks on fleets, by leveraging the inherent deficiencies of road networks identified by techniques from graph analysis. Experimental results on ten cities using realworld courier traces shows that most cities will require upgraded infrastructure to defend driverless vehicles against denial-ofservice attacks. We found several hidden costs that impact equipment designers and operators of driverless vehicles -not least, that road-networks need to be redesigned for robustness against attacks thus raising some fundamental questions about the benefits.